Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

ByCrossFit July 21, 2019

This 2019 trial evaluated the impact of processed food in the diet on caloric intake, independent of other factors.

Twenty subjects were housed in the NIH Clinical Center for four weeks. For two weeks, they ate a diet consisting almost entirely of highly processed foods (i.e., packaged foods available at a typical grocery store). For the remaining two weeks, they ate an unprocessed diet. The order of the diets was randomized for each subject. Besides the level of food processing, these diets were as similar as possible; both contained calorie breakdowns of ~48% carbohydrate, ~37% fat, and 14-16% protein and had similar energy density, sodium, and fiber content. The ultra-processed diet did, however, have higher levels of sugar, higher non-beverage energy density (calories per gram of food), and lower levels of insoluble fiber than the unprocessed diet.

All food was prepared by study staff, and subjects were monitored 24/7. Subjects were fed three daily meals and had unlimited access to snacks. The primary outcome of the study was the number of calories subjects chose to eat under these conditions.

Subjects on the ultra-processed diet consumed an average of 2,979 daily calories, while subjects consuming an unprocessed diet consumed 2,470 — that is, subjects consumed an average of 509 calories (508 ± 106 kcal/day; p = 0.0001) per day more on an ultra-processed diet than an unprocessed diet. The number of calories from protein was nearly identical between diets, but subjects consumed 230 additional calories from fat and 281 additional calories from carbohydrate on an ultra-processed diet compared to an unprocessed diet. Energy intake was fairly consistent over the 14-day periods across both diets. These intake data are shown in Figure 2A and 2B below.

This study was designed to test whether processed food inherently leads to greater caloric intake than unprocessed food, independent of other dietary factors — in other words, whether a diet with the same levels of carbohydrates, fats, protein, sugars, and other nutrients will lead to greater calorie intake if those foods are processed. The perspective that caloric intake may increase as a result of food processing contrasts with a nutrient-centric view that argues processed foods are associated with obesity for reasons explained by their nutrient profile — i.e., they’re higher in carbohydrates, fats and sugars, lower in protein and fiber, etc. The authors concluded the study’s results were consistent with the former perspective, and two otherwise identical diets differing only in the degree of food processing led to significantly different levels of calorie intake.

In a comment also published by the journal, Dr. David Ludwig et al. caution against overinterpreting these results. They note that despite attempts to match the composition of the processed and unprocessed diets, subjects eating a processed diet actually ate more carbohydrate, sugar, saturated fat, and sodium, and consumed a diet with much higher energy density (i.e., more calories per gram of food) than subjects eating an unprocessed diet. Previous data has shown diets with higher energy density increase short-term but not long-term caloric intake, so the observed increase in calories consumed may not hold over weeks to months. They argue it is premature to use these results to support any health claims until they are replicated in longer-term studies.

In the final comment reviewed here, lead author Kevin Hall responds to Ludwig et al.’s critique, correcting some factual errors in the previous response while arguing short-term studies, while flawed, are able to precisely measure food intake and its metabolic consequences in a way longer-term studies cannot. He argues long-term observational research has consistently linked ultra-processed diets to poor health outcomes and reducing processed food intake ought to be a universal goal.

Comments on Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain


Comment thread URL copied!
Back to 190722
Richard Feinman
July 23rd, 2019 at 1:37 pm
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

You have to ask what this is all about. The justification, explicit or implicit is that emphasis on biological science targeting nutrients has failed us and we need to look at something else like processing. But, of course, it hasn't failed us. The science has consistently shown that our original emphasis on fat was incorrect and that dietary carbohydrate was the major controller of metabolic disease. The recent publications from Virta and others establish that low-carbohydrate/keto diets offer a virtual cure for type 2 diabetes and a dramatic adjunct to drugs in type 1. (Diabetes represents the essential metabolic disruption but nothing is better for weight loss).

Low-carb/keto is the best? For the medical establishment this was a grosse neue Shitstorm, as they say in German. How could they deal with such an outcome? Well, you could attack researchers and clinicians personally as in the Tim Noakes, Gary Fettke, Jennifer Elliott, Annika Dahlqvist cases (among others). You could recruit a named Professor of Cardiology to author an epidemiological study showing the low-carb will kill you, never mind that no low-carb diets were studied. Or you could come up with a new angle, red meat bad, plant-based good or ...processing. All have the advantage that you can frequently drag fat back in. Processing is especially good in that you make stuff up to fit your goal. Here's the question: Pick the ultra-processed food:

A. whole milk

B. 2 % milk

C. both

D. neither

Comment URL copied!
Richard Feinman
July 23rd, 2019 at 7:06 pm

Let me be clear. I do not mean to imply any malfeasance on the part of the authors. They did the work and submitted it for peer review.

I think fault attaches to the editors for obviously not getting real peer review. Editors are supposed to recognize controversial subjects and appoint reviewers from both sides of the controversy. This was clearly not done here and constitutes de facto bias.

Medical nutrition at large bears blame for the failure to appreciate how far we have come on focus on nutrients and that those deserve development before turning to more poorly defined and questionable causes. Not that we know the answer. Not that low/carb-keto is any kind of cure-all but that, recent studies on diabetes among others are sufficiently compelling that they deserve serious attention -- rebuttal if appropriate. Choosing to ignore these studies is part of the breakdown in academic medicine that has brought about the consistent complaints about the medical literature for its poor reliability and reproducibility, frequently by the perps themselves.

Comment URL copied!
July 23rd, 2019 at 10:08 am
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

The question of food processing is an important one, no less so because it lacks good definition - the NOVA system is woefully inadequate.

in addition to failing to define their prime variable the study authors didn't design the experiment to illuminate which variables in processing predispose people to fat gain - it just observed what was already known (i.e. people will eat more food when the food is crappy fast food, a mix of flour, sugar and seed oils). this observation is already well established.

Comment URL copied!
Richard Feinman
July 23rd, 2019 at 12:12 pm

My point was that it doesn't even do that because of the lack of conformity of the statistics to Hill's criteria or plain common sense. Also, there is the completely arbitrary nature of the definitions ....frozen green beans are not processed but canned beans are not only processed but "ultra-processed." but the second law of thermodynamics tells you there is more processing in making something cold than making it hot. All of haute cuisine -- or any cuisine -- is out the window because somebody gained 1 kg....remember, that's what it's about, 1 kg.

Comment URL copied!
Richard Feinman
July 22nd, 2019 at 11:57 pm
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

I am surprised that this study is being taken seriously. The key variable “ultra-processed food” is never defined nor can one guess exactly what it might be from descriptions in this and NOVA.

Even assuming one knew what was being studied the changes are distinctly small. This study is a randomized control tria but the conclusion is an association between processing and weight gain although the authors claim "cause." Bradford Hill’s criteria remains a widely accepted test of the extent to which the “cause” in the title is reasonable. The first of Hill’s criteria is the magnitude of the effect. While there is no mathematical or statistical way to say what is large, in this case we all have experience of weight loss and we know that you can lose one kilogram of weight overnight. This means that the changes found in the experiment could be reversed by small changes in behavior and the outcome is not predictable and might change if the experiment were carried out for another day.

The disclaimer is usually that small differences could become important if they were scaled up to the whole population. The question then becomes whether you are dealing with an accurate but small number or just a number that is simply random. Here, you need to look at the error. There are several ways to indicate the spread in the experimental values. The standard deviation (SD) is represented in the familiar bell-shaped curve and gives you a sense of how close measured values are to the mean. One SD on both sides of the mean contains about 65 % of the values. A large SD means big spread and relatively unreliable mean. Small SD indicates better accuracy. The standard error of the mean (SE), however, tells you about the mathematical significance but can be misleading if you are interested in knowing where the experimental data are. The SE always makes your data appear as if it had less error and generally is not the appropriate parameter for understanding reliability of the data. To convert SE to SD, you multiply by the square root of n. For each of the diets reported here that is sqrt (10) or about 3. So the large errors shown in the paper are actually even larger meaning that there will be substantial overlap between the two diet protocols and it may be that there is little reliable difference. The corrected value for weight changes are 0.9 + 0.9 ± 0.95 kg for ultra-processed and -0.9 ± 0.95 kg for unprocessed which underwhelms. More generally almost all the statistical measures in the paper are group statistics which hide the individual changes.

The whole idea is to show that the controlling variable was the processing. In other words, how much can nutrient effects be excluded. So how did it come out?

The paper says “The increased energy intake during the ultra-processed diet resulted from consuming greater quantities of carbohydrate (280 ± 54 kcal/day) and fat (230 ± 53 kcal/day)…”

And: “While we attempted to match several nutritional parameters between the diets, the ultra-processed versus unprocessed meals differed substantially in the proportion of added to total sugar (∼54% versus 1%, respectively), insoluble to total fiber (∼16% versus 77%, respectively), saturated to total fat (∼34% versus 19%), and the ratio of omega-6 to omega-3 fatty acids (∼11:1 versus 5:1).”

On social media such self-serving lack of logic is labelled "and Jesus wept."

And sometimes ultra-processing is good. The Graphic Abstract shows a hamburger as representing processing. I recommend avoiding grinding but rather chopping in a food processor. Julia Child recommended two perfectly balanced chef's knives. That's real ultra but I don't think it will make you fat.

Comment URL copied!
Veronique Oomen
July 22nd, 2019 at 3:27 pm
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

Interesting that the subjects ate exactly the same amount of protein in both cases, while eating far less carbs and fat of unprocessed foods.

Comment URL copied!
Clarke Read
July 22nd, 2019 at 8:32 pm

There is some exceptionally interesting research by Stephen Simpson (out of AUS / UK) looking at this exact observation in more detail. His "protein leverage" hypothesis has long argued we naturally adjust our diets to get a consistent protein intake, with the intake of other nutrients (and total calories) scaling up or down as necessary to ensure protein sufficiency. This hypothesis would be supported by the current data - i.e., that the less protein-rich a diet is, the more of it we might tend to eat. Of course, his is just one of many attitudes toward this question.

Comment URL copied!
Clarke Read
July 22nd, 2019 at 4:52 am
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

This study provides the most direct test to date of a critical , unanswered question in nutrition: Are processed foods inherently bad (i.e., because they are processed), or are processed foods bad simply because they tend to be bad in other ways that are not the direct result of processing (i.e., because they are higher in carbohydrates, or sugars, or more energy dense, or have a higher glycemic index, etc.)

This trial specifically tests at the first claim with respect to obesity. If two diets are as close to identical as possible, except one is highly processed and one is not, does the former lead us to consume more? That is, does processing per se lead to greater caloric consumption when we’re allowed to eat freely?

The authors conclude these data point to “yes”, and strongly so - patients consumed an extra 500 calories per day on the processed food diet. Ludwig et al note why, even taking these results at face value, they might not be meaningful. First off, “as close to identical as possible” may not be close enough to isolate the effect of food processing. The fiber sources differed between diets, and there was more sugar (about 60g / d more) in the processed diet than the unprocessed diet. Total sugar consumption, however, did not differ between diets, which would suggest subjects on the unprocessed diet biased their food choices at least slightly toward the more sugary foods that were available. Ludwig et al focus on the difference in energy density - to reach 2000 calories, not counting beverage calories, subjects eating an unprocessed diet needed to consume nearly twice as much food, by weight (1737g vs 931g). It’s impossible to tell from this study alone whether any of these admittedly small differences are significant and could drive the shocking 500 kcal/d increase in food consumption - but a broader survey of the literature could point us in the right direction. Ludwig et al suggest other short-term studies point to energy density as a cause.

Beyond this, it’s worth questioning whether these results are meaningful given the highly artificial nature of the study. To the authors’ credit (and as reviewed in yesterday’s NYT article), the diets were representative of what a typical highly-processed or unprocessed diet might look like - grocery fare for the former, lightly-prepared whole foods for the latter. But we can’t get past the fact that subjects were kept in a metabolic ward (and for one day each week in a metabolic chamber - basically a prison cell) for two weeks, with snacks readily available all day and more food than they could eat served to them at meals. This might be the best way to test certain laboratory measures of appetite, but it doesn’t match how we eat food in the real world. Ludwig et al imply as much by noting how those same energy density studies have failed to replicate over longer periods. Hall’s study itself even suggests the extent to which environment can affect eating behavior - when subjects were in the metabolic chambers (1 day per week, highly constrained environment), they ate the same number of calories whether they were on an unprocessed or processed diet, while when they were in the ward (6 days per week, moderately constrained environment), they ate much more when the food was processed. How else might these results change when differences in food availability and environment change? Hall argues their study brings a precision that a more loosely controlled study could not, and I’m highly sympathetic to this view, but our tools to deliver and monitor real-world interventions have improved dramatically over the past couple years, and this argument is less compelling than it once was given viable alternatives.

Finally, I do think something of metabolic significance happened to subjects on the processed diet. Urinary C-peptide excretion, a marker of whole-day insulin secretion, was 25% higher on the processed food diet than the unprocessed diet. This was glossed over because it was not statistically significant, albeit at p=0.052. While fasting insulin and glucose and insulin resistance did not differ between groups, we wouldn’t expect a moderate dietary intervention over 2 weeks to lead to changes in these markers. Subjects eating unprocessed diets actually lost weight (more of it fat-free than fat mass), suggesting this diet may have been an improvement over their baseline diet. So what was it about the processed diet, supposedly equal in both carbohydrate and sugars content to the unprocessed diet, that could have led to this increase in insulin?

Hall et al deserve some praise for attempting to isolate and test the impact of a particular dietary feature (food processing) that many of us have set-in-stone opinions about. But this study still has a couple too many asterisks to fully convince me “processed food” is not still just a proxy for some other factors more directly related to the composition of the foods, particularly when it comes to real-world consumption and behavior. That may just be a reflection of my bias.

Comment URL copied!
Clarke Read
July 22nd, 2019 at 6:02 am

A little "spirit of the stairs"...

What we really need is a more clinically relevant understanding of "processed food." We know, from this study and many others, that most diets high in processed foods are worse for us than most diets without them. But why? Is processed food inherently bad, and so the only way to mitigate its effects by eliminating it from the diet? Or is processed food contextually bad - bad because of some less unique factor it does or does not have more often than unprocessed food? Are its differentiated effects on the body entirely cerebral, or do they affect hormones and the metabolism in predictable ways? Is the difference categorical or circumstantial?

We know that the rise of the metabolic disease and obesity epidemics has paralleled the rise of processed food, and so it is very likely that a major contributor to these diseases is something common in processed foods and uncommon in other foods. This study reinforces our understanding that processed foods are, at least in terms of obesity, and at least when consumed within a moderate-carb, moderate-fat diet, worse for us than unprocessed foods. It doesn't get us that much closer to understanding why, except by maybe taking some arguments off the table.

I remain reluctant to believe processed foods are inherently worse for us, for reasons that can't be explained using more mundane or universal nutritional factors. That said, for the individual with the means and capabilities, this does provide further support for a belief we scarcely need more support for - that eating a less-processed diet will, for most of us, most of the time, be a healthier choice.

Comment URL copied!
Mary Dan Eades
July 22nd, 2019 at 4:15 am
Commented on: Study, Comment, & Response: Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain

A not actually very surprising result really. The ultra-processed diet, true to its word, consisted chiefly packaged foods that almost exclusively draw their calories from the same four sources: 1) processed wheat flour 2) processed corn flour 3) industrially processed seed oils, and 4) sugar or HFCS. These players get shaped into a variety of different forms, but basically whether it's ravioli or a bagel or Eggo pancakes, it's some combination of those four ingredients.

The impact of such a diet, high in processed starches, bad fats, and concentrated/refined sugar, is to shunt the fuel partitioning pathways at the sub-cellular level (ie, the electron transport chain) into 'shove the calories into storage' overdrive. That shunting results in relative starvation at the cellular level (well, all the cells but the fat cells) and the need to take in more energy to combat the cellular hunger. That it was nearly 500 kcal a day more in so short a time frame is amazing, considering the tastiness and variety of the unprocessed meals presented and the utter wretchedness of the ultra-processed choices. Additionally, there's the whole issue of the effect that pulverizing food (aka processing) has vis a vis the gut's incretin response, which is a whole other topic that I'm sure will get a thorough going over here before this comment string has wrapped up. In sum, being driven by cellular hunger to consume 500 kcal more of refined calories a day is a pretty sure recipe for gaining weight.

Comment URL copied!